DE2805705A1 - Data communications network linking processors via single highway - functions by each processor sending own address and address of next processor due for access - Google Patents

Abstract

The data network links several data processors via a shared data highway. Each message states the sender and the type of message and is sent to all processors. All processors acknowledge the correct receipt of the message, check bits being included in the message. The messages are sent as blocks of constant length or a multiple of such a length. When no messages are being sent the processors are polled cyclically, whereby each is given access to the data highway. Polling occurs by each processor emitting a data packet containing its own address and the address of the processor next due for access to the highway.

Description

Data exchange procedure between several processors

The invention relates to a data exchange method between several Processors that are connected to one another by common data rails the organizational measures required for the sequential flow of data exchange are required between any partners, evenly distributed to all partners are.

There are processor network operations with data exchange via common Known data rails in which the release of access requests for communication is controlled from a control center, with time-consuming query and response telegrams exchanged or additional signal wires are required for the useful information to convey. The disadvantage of the centrally controlled communication systems in terms of Sometimes they fail if the head office fails altogether The use of redundant control centers is fought, but if the control center fails, they are required switching mechanisms problematic; since the switchover itself is prone to failure and messages can be lost when switching over.

Furthermore, composite processor systems are known in which the participants are connected by ring-shaped or loop-shaped data rails and where the organizational necessary for the control of a sequentially running data traffic Functions are decentralized, i.e. evenly distributed among all participants are; however, the known methods do not meet the condition that a message from to any participant, protected against errors and loss, to a selectable number of others Participants can be dropped off at the same time. However, this condition must be in those data systems are fulfilled in which not only the collection of data is decentralized should take place, but also evaluations of one and the same message sometimes in several processors should run at the same time.

The invention is based on the disadvantages of the known Avoid systems and create a method that is similar to that used in networked processor systems made high demands on availability, reliability and efficiency of the system functions with a precalculable gradual reduction of these properties fulfilled in the event of component failure. This is the case with a data exchange process of the genus mentioned at the beginning achieved through the combination of the following features: a) Each message contains information about the sender and the type of message and is sent to all partners with no errors. All partners confirm the error-free Receiving a message. The receiving partners decide which messages for they are intended.

b) Messages are in the form of blocks of constant length or transmitted to an integer multiple of this constant length.

c) In the idle state, i.e. when there are no messages, occurs on common data rails on which all processors involved in data exchange lie, a cyclical circulation of the access releases for communication from partners to partner by sending a partner who has gained access in a telegram his own Sends out the address and the address of the next partner, who should be granted access.

d) The regularity of the chosen, access to the communication system Redundancy underlying organizational functions is used as a means of security exploited against transmission errors.

The high degree of availability is achieved in the process according to the invention achieved through decentralization of intelligence, such that those functions that of decisive importance for the overall function are (viral functions ") redundant designed, i.e. several processors are able to perform these functions.

If a message is to be sent, the relevant partner waits initially until the next entry. As soon as this is won, the sending begins Station with the same access transfer procedure as in the idle state and adds the message to be sent. A data field is reserved in the header of the message, that designates the type of information, and at the end the message contains the test part, which protects the message against transmission errors.

The partner's receiving facilities contain selection criteria, through which you can decide which types of information from which senders are to be selected for forwarding to the processors.

The selection criteria of the receiving device are under program control of the processors concerned and can be used during initialization or during the In operation. In the access transfer cycle, which is based on the Message transmission, the sending partner expects receipts from everyone Partners.

Every recipient who has received a message without errors acknowledges if: a) a decision was made based on the selection criteria, the message is not forward to the receiving processor, b) decided on the basis of the selection criteria was to forward the message to the receiving processor, and this is ready to receive was.

The acknowledgment is given by a special end character in the access transfer telegram marked.

If one of the mentioned conditions is not met, no acknowledgment is given sent, i.e. the access forwarding telegram is accompanied by a normal end character closed.

In addition, the receiving station sets an error criterion in this case.

If the acknowledgment cycle contains the acknowledgments from all partners, then is the sending procedure is completed and the sending station gives access, namely even if another message is waiting to be transmitted.

The error strategy is implemented in such a way that if there is no at least an acknowledgment, the news broadcast is repeated with an identifier that the repetition is marked. The repeated message is only sent by those receiving devices evaluated, which registered an error in the previous news broadcast to have. If the transmitting device succeeds after a predetermined number of Repetitions fail to receive all receipts, it gives the alarm: ~ Send error "to the Processor and passes the access on. This gives the broadcasting facility responsibility to the processor for further handling of the error, which is dependent on the splication away.

The Kanff device contains each partner for configuration monitoring a "target configuration table" containing the addresses of the partners who are available at the end of the system. In another table, the "actual configuration table", it is constantly registered which partners are active.

If the actual configuration shows gaps compared to the target configuration on, the inactive partners are given the opportunity by periodic queries required to take action: per address gap in the ascending order of partner addresses (modulo n, n = total number of partners according to the target configuration), however, only one (Unsuccessful) call to inactive partners used. If the called partner does not answer, the sending partner immediately calls the next partner in the cyclic address sequence that is active according to the actual configuration. Several are missing Partners with neighboring addresses are used in successive access cycles called.

if a partner fails, the sewer device recognizes that one Partner who wants to hand over access to the bus to the failed partner because he despite repeated call does not respond.

The calling partner then revises his actual configuration table, sets the alarm "configuration change" and calls the next one Partner. In the subsequent access cycle, each partner finds out through comparison The difference between the previous actual configuration table and the new actual configuration and-outputs the configuration change alarm to the processor. Further treatment this error is largely application-dependent and falls within the competencies of the processors receiving the configuration change alarms. In the extreme case if a processor determines that it has no partner, the suspicion is obvious, that the own sewer device is defective; in this case the processor gives the instruction for bridging the sewer unit. In this state, the sewer device is from the The data rail is decoupled on the sending and receiving sides, but does not interfere with the other traffic.

In this situation, the channel device accepts the instruction ~ self-examination11 from the processor. This instruction causes a check of the channel device functions. If errors are found, the procedure is followed by corresponding alarm messages closed. It is your responsibility to deal with any errors of the processor.

If no error is found, you can proceed as described below initiate initialization.

When the partner's power supply is switched off and also after switching on the supply is the partner in the bridged state, i.e. the partner is both on the sending side as well as on the receiving side, decoupled from the data rail and influenced the rest of the data traffic is not.

The processor gives the instruction to connect the sewer device the data rail. With this, the duct device first releases the reception path, i.e. the channel device listens to see whether data traffic is running: If this is the case, it waits the channel device until it receives a directed address call and will then active by also activating the transmission side and the transfer of access caused.

If the sewer device finds the system inactive, it is an initial switch-on: In order to rule out the initialization conflict in this case, that could arise when several partners are switched on at the same time is waiting each partner a certain time, which depends on his partner address, and begins then with the bus access cycle.

The protection of the message header that controls bus access, must be protected to the same extent against unrecognizable transmission errors like the user data block, because e.g. an unrecognized falsification of the sender address of a message can have the same detrimental consequences for process control can be like an undetectable falsification of user data. But the transmission redundancy Significantly restricted in the message header and thus the efficiency of communication be increased if the redundancy, which is the regularity of the cyclical bus access process is used: The receiving facility of each partner is programmed in such a way that that it only accepts access if 1) the telegram of the access request shows no error 2) the call comes from the same sender as the previous call (or several previous calls).

Furthermore, the receiving device then takes a message as error-free if 1) The message block does not show any errors 2) The access request shows no error in the telegram header 3) The return address is the same as the call address in the previous access request.

The availability requirements or the required transmission speed can show that several transmission paths must be available.

In these cases, the transmission facilities per partner and the connecting data rails are created several times.

In these systems, the processors can have preferential routes for the transmission of messages advance. For example, an application can specify that urgent short messages one way, less urgent messages or longer message chains sent another way. The driver programs that send these messages to the transmission devices, take into account the information in the application programs via preferential routes, as long as the desired transmission routes work. Falls a transmission path, the driver program automatically forwards the messages on those transmission routes that still work.

The invention is based on Figures 1 to 7 for the case of a maximum System capacity of 32 partners and with specific message block length definitions for example explained in more detail.

Fig. 1 shows the bus access forwarding telegram, consisting of a telegram header with the start synchronization character ST, the sender address A, the target address for the transfer of access B. The end character follows the head E. ST and E are characters of 6 bits each, for example, sender address A and destination address B are (the system capacity of 32 partners accordingly) E.g. expressed as 5 bits each, each supplemented by an odd parity bit.

Fig. 2 shows a single message block: It begins with the same telegram header ST, A, B that controls the bus access transfer. That closes the message that is initiated by the synchronization character S. To it it is followed by a data field (1 byte) which designates the type of information ("TYP"). The following 7 bytes are freely selectable user data DT. The subsequent check byte PR monitors the message for errors. The telegram is completed by the end character E.

Fig. 3 shows the acknowledgment telegram that results from the bus access forwarding telegram (Fig. 1) differs only by a special end character.

Fig. 4 describes the timing of a rest cycle RZ for Case that three partners with the addresses A1, A2 and A3 are active on the communication system are connected: Partner P1 sends a telegram by first sending its own Sender address Al and then with B2 the destination address for the Passing on of bus access identifies.

The start synchronization character ST and the end character E form the frame of the telegram. As soon as the partner P2 has received the telegram, transfers this the initiative to send in the appropriate way to parent P3. The partner P3 knows from previous cycles that the partners with the Addresses 4 to 32, which will be used in the final expansion existed, not yet were active because it constantly updates its table with the actual configuration. He gives all inactive participants with the addresses 4 to 32 at any time the possibility of becoming active by assigning one of these addresses to each RZ cycle calls. If the called partner does not answer, access is immediately given to the Partner 1 known from the actual configuration table submitted: received in this way all inactive partners 4 to 32 in successive access transfer cycles the possibility of actively engaging in the communication system.

Fig. 5 shows the telegram sequence for 3 active partners, if the partner P1 and P2 each send a message block: After the message has been sent N1 Partner P1 collects the receipts from active partners P2 and P3, before partner P2 can send the message N2, who in turn receives the acknowledgments from collects the partners P3 and P 1.

6 shows the case of the transmission of a chain of message blocks from partner Pl. The sub-blocks N1, N2, N3 all consist in the example described consists of 8 useful bytes and a check byte and are separated from each other by the synchronization character S separated. The last message block is terminated by the end character E.

FIG. 7 shows the case of a faulty message transmission N1 from Partner 1 to partner 3 (error F): In this case, in the acknowledgment cycle connected to the message transmission, the 'Acknowledgment from' Partner 1 (question mark). Partner Pl repeats the message N1 and identifies this by a special character R: Partner P3, who reported the error in the 1st transmission recognizes, expects this repetition and acknowledges if the repetition does not Contains errors, this time in the subsequent acknowledgment cycle. Partner P2 who received the message received error-free when sending, does not forward the repeated message, however, it acknowledges if the reception was error-free.

Claims (23)

Claims 1. Data exchange method between several processes, which are connected to each other by common data rails, in which the organizational Measures for the sequential flow of data exchange between any Partners are required, are evenly distributed among all partners, are identified by the following features: a) Each message contains information about the sender and the type of message and is sent error-proof to all partners. All Partners confirm the error-free receipt of a message. The receiving partners decide which messages are intended for them are, b) Messages are in the form of blocks of constant length or a transmitted as integer multiples of this constant length, c) in the idle state, i.e. if there are no messages, an on takes place on shared data rails which all processors involved in the data exchange are located, a cyclical cycle the access authorizations for communication from partner to partner, in that a partner, who has gained access, his own address and the address in a telegram of the next partner who is to receive access. d) The regularity of the chosen, access to the communication system Redundancy underlying organizational functions is used as a means of security exploited against transmission errors. 2. The method according to claim 1, characterized by decentralization of intelligence, such that those functions that are crucial for the overall function Meaning are (viral functions) designed redundantly, so that several processors are able to perform these functions. 3. The method according to claim 2, characterized in that as ~, vital Function "the transmission itself serves. 4. The method according to claim 1, 2 or 3, characterized in that when sending a message, the partner concerned initially until the next access waits and that, as soon as this access is gained, the sending station with the same access transfer procedure as in the idle state begins and the one to be sent Attach message. 5. The method according to claim 4, characterized in that in the head a data field is reserved for the message, which specifies the type of information, and the message at the end contains a test part which checks the message against transmission errors protects. 6. The method according to claims 1 to 5, characterized in that the partner's receiving facilities contain selection criteria through which they can decide which Types of information from which senders are to be selected for forwarding to the processors. 7. The method according to claim 6, characterized in that the selection criteria of the receiving facilities under Program control of the processors concerned are available and can be set during initialization or during operation. 8. The method according to claims 6 and 7, characterized in that every recipient who received this message without errors has acknowledged in the cases if a) a decision was made based on the selection criteria, the message is not forward to the receiving processor, b) decided on the basis of the selection criteria became, forward the message to the receiving processor and this is ready to receive was. 9. The method according to claim 8, characterized in that the receipt is implemented by a special end character in the access transfer telegram. 10. The method according to claims 8 and 9, characterized in that if one of the conditions a) or b) is not fulfilled, no acknowledgment is sent, i.e. the access forwarding telegram is terminated with a normal end character and also the receiving station in this case Error criterion sets. 11. The method according to claims 8, 9 and 10, characterized in that that if the acknowledgments are available from all partners in the acknowledgment cycle, the send procedure is completed and the transmitting station gives access, even if another message is waiting to be transmitted. 12. The method according to claims 8 to 11, characterized in that in the absence of at least one acknowledgment, the message broadcast with an identifier is repeated, which marks the repetition, and that the repeated message is only evaluated by those receiving devices that were used in the previous News broadcast registered an error. 13. The method according to claims 8 to 12, characterized in that the sending device if all receipts have not been successfully received according to a predetermined one Number of repetitions sends the alarm "send error" to the processor and the Passes access. 14. The method according to claims 1 to 13, characterized in that the channel device of each partner for configuration monitoring an 11 target configuration table " contains the addresses of the partners that were available during the final system configuration are, and that in another table, the "actual configuration table", permanently it registers which partners are active. 15. The method according to claims 1 to 14, characterized in that when there are gaps in the actual configuration compared to the target configuration the inactive Partners through periodic queries is offered to take action. 16. The method according to claim 15, characterized in that per address gap in the ascending order of the partner addresses (modulo n, n = total number of partners according to the target configuration) only one (unsuccessful) call is used. 17. The method according to claims 15 and 16, characterized in that If the called partner does not answer, the calling partner immediately informs them calls the next partner in the cyclic address sequence according to the actual configuration is active, and that if there are no more <partners with neighboring addresses these can be called in successive access cycles. 18. The method according to claims 15 to 17, characterized in that the failure of a partner is recognized by the sewer device of that partner who wants to hand over access to the bus to the failed partner by the called partner does not answer despite repeated calls, whereupon the calling partner Partner being the actual configuration table revised, the alarm configuration change " and calls the next partner. 19. The method according to claims 14 to 18, characterized in that every partner who notices that the actual configuration has been changed upon receipt of the access release has changed since the last access, sets the alarm ~ configuration change. 20. The method according to claims 17, 18 and 19, characterized in that that-in the extreme case, when a processor receiving the configuration change alarm determines that it has no partner, this processor the instruction to bridge of the sewer device indicates the state in which the sewer device is transmitting and receiving is decoupled from the data rail, but not the rest of the data traffic hindered, and that the channel device in this Zistand the instruction from the processor Self-check accepts which instruction is a check of the sewer unit functions causes. 21. The method according to claims 1 to 19, characterized in that when the power supply to a partner is switched off and also after the power supply has been switched on this partner in the state ?? bridged ?? is that. the duct device on instruction of the processor. first releases the reception path, i.e. determines whether data traffic runs, and that when data traffic is detected, the channel device waits until it is all about itself. receives a directed address call, and then becomes active, by also activating the transmission side and initiating the transfer of access. 22. The method according to claim 20, characterized in that in the case system inactivity detected by the channel device, in which case It is an initial switch-on, each partner a certain one from his Partner address-dependent time waits and then begins with the bus access cycle. 23. The method according to claims 1 to 21, characterized in that the receiving device of each partner is programmed in such a way that it only then Accepts entry, if 1) the telegram of the access request shows no error, 2) the call comes from the same sender as the previous call or also several previous calls and also the receiving device is programmed in such a way that it then records a message as error-free if 1) the message block does not reveal any errors, 2) the access request shows no error in the telegram header, 3) the sender address is the same is like the call address in the previous access request.